On Tue, 2014-01-21 at 23:52 +0100, skodde wrote: > On Tue, Jan 21, 2014 at 11:06 PM, Ben Hutchings <b...@decadent.org.uk> wrote: > > So far as I am aware, the write failure is properly reported (except > > that the operation is retried and the failure logged many times, which > > is also a known bug) and previously written data are readable. If > > that is the case, I don't believe this qualifies as data corruption. > > I'm not sure if every write operation triggers a corruption, but, as > an example, if you try to run a fsck (forcing it if necessary) on an > existing ext4 filesystem on the encrypted volume, the filesystem will > be heavily corrupted.
That sounds like there could be a bug in fsck (writing changes in the wrong order), but I agree that this is also a grave bug in the kernel. Ben. -- Ben Hutchings compatible: Gracefully accepts erroneous data from any source
signature.asc
Description: This is a digitally signed message part